add pytorch repo/ add rocm 6.3.4

Author: mixa3607

checkpoint.sh

@@ -10,35 +10,8 @@ fi
 TAG_NAME=$(git_get_current_tag)
 if [ "$TAG_NAME" == "" ]; then
   TAG_NAME="$(date +%Y%m%d%H%M%S)"
-  TAG_COMMENT="$(
-    echo "Release $TAG_NAME"
-    echo "Checkpoints:"
-    echo "- ROCm: $ROCM_VERSION"
-    echo "- ComfyUI: $COMFYUI_GIT_REF"
-    echo "- llama.cpp: $LLAMA_GIT_REF"
-  )"
-  git tag -a "$TAG_NAME" -m "$TAG_COMMENT"
-  echo -e "New tag $TAG_NAME:\n$TAG_COMMENT"
+  git tag -a "$TAG_NAME"
+  echo -e "New tag $TAG_NAME" -m "none"
 else
   echo "Commit already tagged with $TAG_NAME"
 fi
-
-git push origin "$TAG_NAME"
-#git push origin
-
-pushd rocm
-./build-and-push.rocm.sh
-popd
-
-pushd llama.cpp
-./build-and-push.rocm.sh
-./build-and-push.vulkan.sh
-popd
-
-pushd comfyui
-./build-and-push.comfyui.sh
-popd
-
-pushd vllm
-./build-and-push.vllm.sh
-popd

env.sh

@@ -42,3 +42,4 @@ source $(dirname ${BASH_SOURCE[0]})/rocm/env.sh
 source $(dirname ${BASH_SOURCE[0]})/llama.cpp/env.sh
 source $(dirname ${BASH_SOURCE[0]})/comfyui/env.sh
 source $(dirname ${BASH_SOURCE[0]})/vllm/env.sh
+source $(dirname ${BASH_SOURCE[0]})/pytorch/env.sh

pytorch/build-and-push.torch.sh

@@ -0,0 +1,28 @@
+#/bin/bash
+set -e
+
+cd $(dirname $0)
+source ../env.sh
+
+IMAGE_TAGS=(
+  "$TORCH_IMAGE:${TORCH_VERSION}-rocm-${TORCH_ROCM_VERSION}-${REPO_GIT_REF}"
+  "$TORCH_IMAGE:${TORCH_VERSION}-rocm-${TORCH_ROCM_VERSION}"
+)
+
+if docker_image_pushed ${IMAGE_TAGS[0]}; then
+  echo "${IMAGE_TAGS[0]} already in registry. Skip"
+  exit 0
+fi
+
+DOCKER_EXTRA_ARGS=()
+for (( i=0; i<${#IMAGE_TAGS[@]}; i++ )); do
+  DOCKER_EXTRA_ARGS+=("-t" "${IMAGE_TAGS[$i]}")
+done
+
+mkdir ./logs || true
+docker buildx build ${DOCKER_EXTRA_ARGS[@]} --push \
+  --build-arg BASE_ROCM_IMAGE="${PATCHED_ROCM_IMAGE}:${TORCH_ROCM_VERSION}-complete" \
+  --build-arg ROCM_ARCH="${ROCM_ARCH}" \
+  --build-arg PYTORCH_BRANCH=$TORCH_BRANCH \
+  --build-arg PYTORCH_VISION_BRANCH=$TORCH_VISION_BRANCH \
+  --target final -f ./torch.Dockerfile --progress=plain ./submodules 2>&1 | tee ./logs/build_$(date +%Y%m%d%H%M%S).log

pytorch/env.sh

@@ -0,0 +1,17 @@
+#/bin/bash
+
+pushd $(dirname ${BASH_SOURCE[0]})
+
+# rocm version
+if [ "$TORCH_ROCM_VERSION" == "" ];          then TORCH_ROCM_VERSION=6.3.3; fi
+# torch git checkpoint
+if [ "$TORCH_PYTORCH_BRANCH" == "" ];        then TORCH_PYTORCH_BRANCH="v2.7.1"; fi
+# vision git checkpoint
+if [ "$TORCH_PYTORCH_VISION_BRANCH" == "" ]; then TORCH_PYTORCH_VISION_BRANCH="v0.21.0"; fi
+
+# destination image
+if [ "$TORCH_IMAGE" == "" ]; then
+  TORCH_IMAGE=docker.io/mixa3607/pytorch-gfx906
+fi
+
+popd

pytorch/preset.torch-2.7.1-rocm-6.3.4.sh

@@ -0,0 +1,5 @@
+#!/bin/bash
+
+export TORCH_ROCM_VERSION="6.3.4"
+export TORCH_PYTORCH_BRANCH="v2.7.1"
+export TORCH_PYTORCH_VISION_BRANCH="v0.21.0"

pytorch/readme.md

@@ -0,0 +1,46 @@
+# PyTorch GFX906
+Tensors and Dynamic neural networks in Python with strong GPU acceleration.
+
+Packages:
+- torch
+- torchvision
+
+Recommend use `docker.io/mixa3607/pytorch-gfx906:v2.7.1-rocm-6.3.3`
+
+## Build
+See build vars in `./env.sh`. You also may use presetis `./preset.*.sh`. Exec `./build-and-push.torch.sh`:
+```bash
+$ . preset.torch-2.7.1-rocm-6.3.3.sh
+$ ./build-and-push.torch.sh
+~/REPOS/mixa3607/llama.cpp-gfx906/rocm ~/REPOS/mixa3607/llama.cpp-gfx906/rocm
+~/REPOS/mixa3607/llama.cpp-gfx906/rocm
+~/REPOS/mixa3607/llama.cpp-gfx906/llama.cpp ~/REPOS/mixa3607/llama.cpp-gfx906/rocm
+~/REPOS/mixa3607/llama.cpp-gfx906/rocm
+~/REPOS/mixa3607/llama.cpp-gfx906/comfyui ~/REPOS/mixa3607/llama.cpp-gfx906/rocm
+~/REPOS/mixa3607/llama.cpp-gfx906/rocm
+~/REPOS/mixa3607/llama.cpp-gfx906/vllm ~/REPOS/mixa3607/llama.cpp-gfx906/rocm
+~/REPOS/mixa3607/llama.cpp-gfx906/rocm
+#0 building with "remote" instance using remote driver
+
+#1 [internal] load build definition from rocm.Dockerfile
+#1 transferring dockerfile: 4.95kB done
+#1 DONE 0.0s
+
+#2 [auth] dockerio-proxy/rocm/dev-ubuntu-24.04:pull rocm/dev-ubuntu-24.04:pull token for registry.arkprojects.space
+#2 DONE 0.0s
+
+#3 [internal] load metadata for docker.io/rocm/dev-ubuntu-24.04:7.0-complete
+#3 DONE 1.8s
+
+#4 [internal] load .dockerignore
+#4 transferring context: 2B done
+#...............
+#24 exporting to image
+#24 pushing layers 6.5s done
+#24 pushing manifest for docker.io/mixa3607/rocm-gfx906:7.0.0-20251005035204-complete@sha256:00532f62462e80d51e48b021afb7875af53164455c84dc28b24eb29d39aa0005
+#24 pushing manifest for docker.io/mixa3607/rocm-gfx906:7.0.0-20251005035204-complete@sha256:00532f62462e80d51e48b021afb7875af53164455c84dc28b24eb29d39aa0005 3.3s done
+#24 pushing layers 2.0s done
+#24 pushing manifest for docker.io/mixa3607/rocm-gfx906:7.0.0-complete@sha256:00532f62462e80d51e48b021afb7875af53164455c84dc28b24eb29d39aa0005
+#24 pushing manifest for docker.io/mixa3607/rocm-gfx906:7.0.0-complete@sha256:00532f62462e80d51e48b021afb7875af53164455c84dc28b24eb29d39aa0005 2.2s done
+#24 DONE 17.6s
+```

pytorch/submodules/.gitkeep

@@ -0,0 +1,57 @@
+ARG BASE_ROCM_IMAGE="docker.io/mixa3607/vllm-gfx906:latest"
+ARG ROCM_ARCH="gfx906"
+ARG PYTORCH_REPO="https://github.com/pytorch/pytorch.git"
+ARG PYTORCH_BRANCH="v2.7.1"
+ARG PYTORCH_VISION_REPO="https://github.com/pytorch/vision.git"
+ARG PYTORCH_VISION_BRANCH="v0.21.0"
+
+############# Base image #############
+FROM ${BASE_ROCM_IMAGE} AS rocm_base
+# Install basic utilities and Python 3.12
+RUN apt-get update && apt-get install -y software-properties-common git python3-pip && \
+    add-apt-repository ppa:deadsnakes/ppa && \
+    apt-get update -y && \
+    apt-get install -y python3.12 python3.12-dev python3.12-venv \
+    python3.12-lib2to3 python-is-python3 python3.12-full && \
+    update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.12 1 && \
+    update-alternatives --set python3 /usr/bin/python3.12 && \
+    ln -sf /usr/bin/python3.12-config /usr/bin/python3-config && \
+    python3 -m pip config set global.break-system-packages true && \
+    pip install amdsmi==$(cat /opt/ROCM_VERSION_FULL) && \
+    true
+
+# Set environment variables
+ARG ROCM_ARCH
+ENV ROCM_ARCH=$ROCM_ARCH
+ENV PYTORCH_ROCM_ARCH=$ROCM_ARCH
+ENV PATH=/opt/rocm/llvm/bin:$PATH
+ENV ROCM_PATH=/opt/rocm
+ENV LD_LIBRARY_PATH=/opt/rocm/lib:/usr/local/lib:
+
+############# Build torch + vision #############
+FROM rocm_base AS build_torch
+RUN pip install setuptools wheel packaging cmake ninja setuptools_scm jinja2
+
+ARG PYTORCH_REPO
+ARG PYTORCH_BRANCH
+RUN ls -la /opt/
+RUN git clone --depth 1 --recurse-submodules --shallow-submodules --jobs 4 --branch ${PYTORCH_BRANCH} ${PYTORCH_REPO} pytorch
+RUN cd pytorch && pip install -r requirements.txt
+RUN cd pytorch && \
+    python3 tools/amd_build/build_amd.py && \
+    CMAKE_PREFIX_PATH=$(python3 -c 'import sys; print(sys.prefix)') python3 setup.py bdist_wheel --dist-dir=/dist && \
+    pip install /dist/*.whl
+
+ARG PYTORCH_VISION_REPO
+ARG PYTORCH_VISION_BRANCH
+RUN git clone --depth 1 --recurse-submodules --shallow-submodules --jobs 4 --branch ${PYTORCH_VISION_BRANCH} ${PYTORCH_VISION_REPO} vision
+RUN cd vision && python3 setup.py bdist_wheel --dist-dir=/dist \
+    && pip install /dist/*.whl
+
+############# Install all #############
+FROM rocm_base AS final
+RUN --mount=type=bind,from=build_torch,src=/dist/,target=/dist_torch \
+    pip install /dist_torch/*.whl && \
+    true
+
+CMD ["/bin/bash"]

pytorch/torch.Dockerfile

@@ -0,0 +1,4 @@
+#!/bin/bash
+
+export ROCM_VERSION="6.3.4"
+export ROCM_IMAGE_VER="6.3.4"